Overview

• The aggregation-onset algorithm predicts misfolded protein aggregation events from fluorescence images with 91% accuracy, marking the first real-time forecast of such processes.
• A separate deep learning model detects mature aggregates to trigger Brillouin microscopy for measuring biomechanical properties like elasticity.
• Dynamic switching between fluorescence and Brillouin modalities minimizes fluorescent labeling and preserves native sample biophysics.
• This autonomous system delivers the first dynamic, label-free capture of the biomechanical evolution of protein aggregation as it occurs.
• Nature Communications reports the breakthrough that could open new pathways for precision medicine and drug discovery targeting neurodegenerative disease mechanisms.